Data Preprocessing for Intrusion Detection System Using Encoding and Normalization Approaches
Mahendra Yadav, R. Kalpana · 2019
This work presents a data preparation and data preprocessing framework to support deep learning and network security experts in producing qualitative data for empirical experimental analysis of intrusion detection data. The Onehotencoder and min-max normalization approaches are used in this proposed preprocessing module. This research paper focuses primarily on analyzing two datasets, specifically KDD Cup' 99, and NSL-KDD datasets, which are commonly used to investigate computer network intrusion detection. The database of the KDD Cup ' 99 consist of five million files, each with 41 attributes that can categorize malicious intrusions into four classes: Probe, DoS, U2R and R2L. The real traffic data cannot be replicated by the KDD cup'99 data set because it was produced over a virtual computer network by simulation. Duplicate and obsolete records were omitted from training and testing in the NSL-KDD database from the KDD Cup ' 99 dataset.